enable / disable Long Task instrumentation based on presence of LongTask observer

third_party/WebKit/Source/core/timing/PerformanceObserver.cpp

Index: third_party/WebKit/Source/core/timing/PerformanceObserver.cpp
diff --git a/third_party/WebKit/Source/core/timing/PerformanceObserver.cpp b/third_party/WebKit/Source/core/timing/PerformanceObserver.cpp
index 920a8d5ce3e84a6acd3a792946ff2c0fbc6d8802..374567590d53a8713f1c6a05e4f78a5a48a663cf 100644
--- a/third_party/WebKit/Source/core/timing/PerformanceObserver.cpp
+++ b/third_party/WebKit/Source/core/timing/PerformanceObserver.cpp
@@ -52,14 +52,16 @@ void PerformanceObserver::observe(const PerformanceObserverInit& observerInit, E
         m_performance->updatePerformanceObserverFilterOptions();
     else
         m_performance->registerPerformanceObserver(*this);
+    m_performance->enableLongTaskInstrumentation();

[caseq, 2016/09/19 22:18:06]

Should we rather make this part of
PerformanceBase::registerPerformanceObserver()?

[panicker, 2016/09/20 12:00:39]

Yes, I started with that in this CL, but the issue was observe() didn't
necessarily call PerformanceBase::registerPerformanceObserver().
Though moving it back, now that I've merged enable / disable into 1 method this
is fine.

     m_isRegistered = true;
 }
 
 void PerformanceObserver::disconnect()
 {
-    if (m_performance)
+    if (m_performance) {
         m_performance->unregisterPerformanceObserver(*this);
-
+        m_performance->disableLongTaskInstrumentation();

[caseq, 2016/09/19 22:18:06]

ditto, consider moving to PerformanceBase::unregisterPerformanceObserver()

[panicker, 2016/09/20 12:00:39]

Done.

+    }
     m_performanceEntries.clear();
     m_isRegistered = false;
 }